Questionnaire Design II - Vanderbilt...
Transcript of Questionnaire Design II - Vanderbilt...
Questionnaire Design II
Mario Davidson, PhDVanderbilt University School of Medicine
Department of Biostatistics
Defining the Domain: Class Discussion
Proposed Research Study:
Graduates of Vanderbilt University School of
Medicine (VUSM) are successful. The Research Question Literature Search
Existing Questionnaires
Types of Questions: Open-ended
Examples How would you
describe success? What are the
primary tasks you perform at work?
Advantages Disadvantages
Types of Questions: Closed
Examples
1.What is your annual salary?
2.On a scale of 1 - 10, where 1 is dislike and 10 is like, rate how well you like your job.
Advantages Disadvantages
Types of Questions: Semi-closed
Example
1.What role have you taken the most in published research?
− Author− Reviewer− Other (specify)
____________− None− Don't Know
Defining the Domain: Brainstorming Session for
QuestionsIn small groups on a sheet of paper, write one
open, closed, and semi-closed questions and the
choices of answers for graduates of VUSM that
would help us understand the proposed research
study: Graduates of Vanderbilt University
School of Medicine have a strong
impact on published research.
Vocabulary
Appropriate Language Slang Ethnic Regional
Familiar Language
Wuz up?
Dot gone
Y'allI re
con'
Fixin' to
Cocodrie
Pro
ps
Formulating the Questions: Class Discussion
Discuss the following questions requested to be answered by
graduates of Vanderbilt University School of Medicine:
a. Are you an editor of a journal or a reviewer?
b. Have you recently had an article published?
c. Don't you believe receiving grant money is the
most important aspect of successful research?
d. Out of the published research that you have contributed to, what was the highest impact factor?
e. How often have you bribed someone in order to be published?
Formulating the Questions: Class Discussion on Biases
a. Are you an editor of a journal or a reviewer? Double-
barrelled
Solution: Make two separate questions.
b. Have you recently had an article published? Vague words
Solution: Try different words and ask others for feedback.
c. Don't you believe receiving grant money is the most
important aspect of successful research? Leading
Solution: Reword the question. On a scale of 1-10...
Formulating the Questions: Class Discussion on Biases
d. Out of the published research that you have contributed to, what was the highest impact factor? Ambiguous/Technical jargon/Relies on memory
Solution: Reword. Try not to rely too much on recall.
e. How often have you bribed someone in order to be published? Sensitive
Solution: If you must ask sensitive questions, try giving a lead sentence such as: Some researchers have admitted buying lunch, sending a gift.....
Formulating the Questions: Other Biases
Complex
Insensitive measure: Scale of 1 – 2 – 3 vs. 1 – 10
Too few categories
Missing and/or overlapping intervals
Horizontal vs. vertical response format
Length of questionnaire
Placement of questions
Skipping questions and branching
Cultural differences
Avoid hypothetical questions
Formulating the Questions: Other Biases (Dillman, 2007)
Do not use, check-all-that-apply question formats
Choose question wordings that are comparable to previous collected data
Consider questionnaires as conversations. Do not switch topics often. It gives the appearance of not listening to the responder.
Choose the first question carefully. It should apply to everyone and be easy and interesting.
Ask one question at a time. Ex: Moral Distress frequency and occurrence
Use spacing
Use bold for questions and light print for answer choices
Validity and Reliability Reliability
Test-Retest Method Equivalent Form Internal Consistency
Validity Internal and External Face Content Criterion Construct
Reliability and Validity
There are different definitions for the types of reliability and validityReliability refers to the consistency of a set of measurements.Validity refers to what degree the research reflects the given research problem.A measure may not be valid if it is not reliable.
http://www.experiment-resources.com/research-methodology.html
Reliability: Test-Retest
The same instrument is provided to the same subjects twice.
The instrument should produce similar results. An issue with this method is finding the appropriate
time to re administer the instrument. Ex: If a pre-survey is administered regarding knowledge
of HIV before an intervention workshop, the post-survey will most likely give very different results. However if a post survey is given and the same survey is given in three weeks, you would expect the results to be similar.
Reliability: Equivalent Form
Different versions of the instrument are created
Results should be similar
Reliability: Internal Consistency
Items on the measure should provide consistent scores.
Split-halves Test Compare the odd questions to the evens Should correlate well
Validity: Internal and External
Internal Measures the confidence in which there is a cause-effect
relationship. “Could there be an alternative cause, or causes, that
explain my observations and results?”
External Can the results be generalizable to a larger population “To what populations, setting, treatment variables and
measurement variables can this effect be generalized.”
http://www.experiment-resources.com/external-validity.html
Validity: Face
At “face value” does the instrument make sense. Does not require expert opinion. Arguably the weakest type of validity Sometimes it is difficult to measure or get experts
opinion Ex of Face Validity: A survey may be given to
patients with irritable bowel syndrome regarding the syndrome. They may be asked to comment on if the relevant questions were asked.
Validity: Content
How well does the instrument represent every characteristic of the construct
Usually requires expert opinions
Statistical analysis should be considered
Validity: Criterion
Criterion Validity: How well does a measurement estimate or predict certain abilities
Concurrent
Predictive
The major differences deal with the time of administration
Validity: Concurrent
Concurrent Validity – How well does a measure correlate with another measure of the same type?
Concurrent suggest that the measures should be given (approximately) at the same time.
Ex: If two different IQ tests are given, they should correlate well.
Validity: Predictive
Predictive Validity – How well does a measure predict a construct in the future?
Ex: A survey may be given to understand if medical students receiving this lecture will be inclined to create a survey later in their careers. We may see if there is a correlation with the number of surveys designs they participate in once they have worked as a physician for 10 years.
Validity: Construct
Construct Validity – How well does the construct measure what it is suppose to be measuring?
Arguably the most important validity and difficult to prove.
Ex: A survey measuring medical students' wellness. Wellness is an abstract, theoretical construct.
If you can prove convergence and/or divergence, this may be enough.
Validity: Convergent and Divergent
Convergent Validity – Do constructs relate to expected constructs they should be related to?
Divergent Validity – Is the construct unrelated to construct believed to be unrelated?
Validity: Convergent
The construct, Self-Esteem, should relate with all other constructs. Self-worth and confidence should relate and social skills and self appraisal should too.
http://www.experiment-resources.com/convergent-validity.html
Preparation
Type of Analysis Establishment of Codes
Open-ended Questions Closed Questions
Code Book Live Document Precision and Completion Record of Algorithms
Preparation
Clearly define the purpose of the study. Provide clear and concise instructions. Discuss the policy on confidentiality Discuss the person's right to refuse any question(s). Include identifying data on multi-page
questionnaires.
The Pretest
Purpose
Appropriate Wording and Questions
Blank Answers Right Sample Skip Patterns Read Body Language Interpretations Similar
Length Request Feedback from
Subjects Listen to Feedback Software (Zoomerang,
Adobe Live Cycle Designer, SurveyMonkey.com, REDCap Survey)
Considerations
Pilot Test
The pilot test is a formalized small scale administration
Some Questions How is the response rate? Are there questions that are not being answered? Is relevant information being obtained from open-ended
questions? How much variability is occurring for each item?
Conclusion Start as Soon as Possible (ASAP)!
Don't Recreate the Wheel!
Pretest and Pilot the Right Population
Review, Revise, and Test
ReferencesDillman, D.A. (2007). Mail and internet surveys: The tailored design method(2nd ed.).
Hoboken, NJ: John Wiley & Sons, Inc.
http://www.idrc.ca/openebooks/069-1/img/designandco_144_la_0.jpg
http://www.experiment-resources.com/research-methodology.html
http://www.experiment-resources.com/external-validity.html
http://www.experiment-resources.com/convergent-validity.html